Hybrid fuzzy closed-loop sub-micron critical dimension control in wafer manufacturing

• US 6,556,876 B1
• Filed: 10/12/2000
• Issued: 04/29/2003
• Est. Priority Date: 10/12/2000
• Status: Active Grant

• Current Assignee: National Semiconductor Corporation (Texas Instruments, Inc.)
• Original Assignee: National Semiconductor Corporation (Texas Instruments, Inc.)
• Inventors: Prosack, Hank, Swan, Kenneth, Ferland, Patrick H.
• Primary Examiner(s): Follansbee, John
• Assistant Examiner(s): Hartman, Jr., Ronald D
• Application Number:
  US09/689,008
• Time in Patent Office: 929 Days
• Field of Search: 700/50, 700/121, 700/45-46, 700/51-52, 706/47, 706/52, 706/900, 706/2
• US Class Current: 700/50
• CPC Class Codes: G05B 19/41875 characterised by quality surveillance of production
• CPC Class Codes: G05B 2219/34065 Fuzzy logic, controller
• CPC Class Codes: G05B 2219/42136 Fuzzy feedback adapts parameters model
• CPC Class Codes: G05B 2219/45031 Manufacturing semiconductor wafers
• CPC Class Codes: H01L 22/20 Sequence of activities consisting of a plurality of measurements, corrections, marking or sorting steps
• CPC Class Codes: Y02P 80/30 Reducing waste in manufacturing processes; Calculations of released waste quantities
• CPC Class Codes: Y02P 90/02 Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]

Abstract

Variation in results of a semiconductor fabrication process is minimized by adjusting process parameters utilizing a fuzzy-controlled learning feedback system. The fuzzy-controlled learning system receives as inputs the results of the fabrication process and then converts these results into a fuzzy set defined by a membership function. An inference engine applies a fuzzy rule base to map an output fuzzy set from the input fuzzy set. The fuzzy output set is then converted to crisp outputs which automatically adjusts parameters of the fabrication process in order to minimize variation.

Claims

1. A method of controlling variation in a process utilizing a fuzzy-controlled learning feedback system, the method comprising:

generating a plurality of crisp process input values related to control of the process, the crisp process input values including at least one crisp process input value related to a specific process parameter value to be controlled;

determining a target bias value suitable for converting said at least one crisp process value input value to a final process value that results in a desired final specific process parameter value;

determining an amount of variation needed in a preselected step of the process to obtain the target bias value;

performing the preselected step of the process in accordance with the determined amount of variation, thereby generating a new at least one crisp process input value;

processing the new at least one crisp process input value utilizing a fuzzy feedback system to generate an adjustment value;

utilizing the adjustment value to determine an updated target bias value; and

utilizing the updated target bias value to iteratively repeat the foregoing recited steps of the process until the desired final specific process parameter value is substantially obtained.

2. A method of controlling variation in a polysilicon etching process utilizing a fuzzy-controlled learning feedback system, the method comprising:

receiving a plurality of crisp process input variables, including a product identification input variable which determines a precise expected correlation between etch bias and bottom anti-reflective coating over-etch (BARC OE), and further including an average develop inspect critical dimension (DICD) input variable of an incoming wafer lot to be processed and a final inspect critical dimension (FICD) input variable of wafer lots already processed;

utilizing the DICD input variable and the FICD Target input variable to calculate a target bias value necessary to convert the incoming DICD into the desired FICD Target;

determining an amount of BARC overetching necessary to obtain the target bias value;

etching a layer of polysilicon in accordance with the DICD input variable, the FICD Target input variable and the target-bias value;

measuring the FICD value actually resulting from the preceding etching step;

feeding the measured FICD value back to a feedback loop that utilizes a fuzzy feedback system to generate at least one fuzzy output variable;

utilizing the at least one fuzzy output variable to iteratively repeat the foregoing recited steps of the process until the actual FICD value measured on the wafer lot is substantially the same as the FICD Target input variable.